I would be skeptical that the derived attitude indications based on GPS
course, speed and altitude changes would be
sufficient in an upset. A vacuum failure can be insidious and the airplane
will likely be in an unusual attitude
before you deduce the problem. I would be curious if anyone has
experimented with this in earnest.

The MTBF of a dry pump may be as low as 400 hours. I believe the backup
electrical AI can be used as a replacement
for your turn-and-bank indicator. I do not have one, but it would afford a
tremendous piece of mind and I am convinced
your scan would pick up the inconsistency sooner than the cross check
afforded by the standard six pack. Our aircraft have
an electric backup vacuum source (which would require you to recognize the
failure and flip the switch). The electric
source does not generate the minimum 4.5lbs/in/in but I have never
investigated how this might affect the instrument
performance.

Regards,
Todd